1 files changed, 79 insertions, 0 deletions
diff --git a/src/alfg/add_lagg_fibonacci_prng.c b/src/alfg/add_lagg_fibonacci_prng.c
new file mode 100644
index 0000000..e84f90e
--- /dev/null
+++ b/src/alfg/add_lagg_fibonacci_prng.c
@@ -0,0 +1,79 @@
+/*
+ * Additive Lagged Fibonacci Generator (ALFG) Implementation
+ * Author: Fabian Druschke
+ * Date: 2024-03-13
+ *
+ * This is an implementation of the Additive Lagged Fibonacci Generator (ALFG),
+ * a pseudorandom number generator known for its simplicity and good statistical properties
+ * for a wide range of applications. ALFGs are particularly noted for their long periods
+ * and efficiency in generating sequences of random numbers. However, like many other PRNGs,
+ * they are not suitable for cryptographic purposes due to their predictability.
+ *
+ * As the author of this implementation, I, Fabian Druschke, hereby release this work into
+ * the public domain. I dedicate any and all copyright interest in this work to the public
+ * domain, making it free to use for anyone for any purpose without any conditions, unless
+ * such conditions are required by law.
+ *
+ * This software is provided "as is", without warranty of any kind, express or implied,
+ * including but not limited to the warranties of merchantability, fitness for a particular
+ * purpose, and noninfringement. In no event shall the authors be liable for any claim,
+ * damages, or other liability, whether in an action of contract, tort, or otherwise, arising
+ * from, out of, or in connection with the software or the use or other dealings in the software.
+ *
+ * Note: This implementation is designed for non-cryptographic applications and should not be
+ * used where cryptographic security is required.
+ */
+
+#include "add_lagg_fibonacci_prng.h"
+#include <stdint.h>
+#include <string.h>
+
+#define STATE_SIZE 64  // Size of the state array, sufficient for a high period
+#define LAG_BIG 55  // Large lag, e.g., 55
+#define LAG_SMALL 24  // Small lag, e.g., 24
+#define MODULUS ( 1ULL << 48 )  // Modulus for the operations, here 2^48 for simple handling
+
+void add_lagg_fibonacci_init( add_lagg_fibonacci_state_t* state, uint64_t init_key[], unsigned long key_length )
+{
+    // Simple initialization: Fill the state with the key values and then with a linear combination of them
+    for( unsigned long i = 0; i < STATE_SIZE; i++ )
+    {
+        if( i < key_length )
+        {
+            state->s[i] = init_key[i];
+        }
+        else
+        {
+            // Simple method to generate further state values. Should be improved for serious applications.
+            state->s[i] = ( 6364136223846793005ULL * state->s[i - 1] + 1 ) % MODULUS;
+        }
+    }
+    state->index = 0;  // Initialize the index for the first generation
+}
+
+void add_lagg_fibonacci_genrand_uint256_to_buf( add_lagg_fibonacci_state_t* state, unsigned char* bufpos )
+{
+    uint64_t* buf_as_uint64 = (uint64_t*) bufpos;  // Interprets bufpos as a uint64_t array for direct assignment
+    int64_t result;  // Use signed integer to handle potential negative results from subtraction
+
+    for (int i = 0; i < 4; i++) {
+        // Subtract the two previous numbers in the sequence
+        result = (int64_t)state->s[(state->index + LAG_BIG) % STATE_SIZE] - (int64_t)state->s[(state->index + LAG_SMALL) % STATE_SIZE];
+
+        // Handle borrow if result is negative
+        if (result < 0) {
+            result += MODULUS;
+            // Optionally set a borrow flag or adjust the next operation based on borrow logic
+        }
+
+        // Store the result (after adjustment) back into the state, ensuring it's positive and within range
+        state->s[state->index] = (uint64_t)result;
+
+        // Write the result into buf_as_uint64
+        buf_as_uint64[i] = state->s[state->index];
+
+        // Update the index for the next round
+        state->index = (state->index + 1) % STATE_SIZE;
+    }
+}
+